Governed OpenClaw: the power of an open agent, with the guardrails an enterprise needs.
Open-source agents that run on your own machine, use any model, and learn as you use them. This page covers OpenClaw, Hermes Agent, as of October 2026.
Clone, side by side
Pick a product. We'll tell you where it's strong, too.
Governed OpenClaw: the power of an open agent, with the guardrails an enterprise needs.
Their pitch. Open-source agents that run on your own machine, use any model, and learn as you use them.
Where they're strong
- Very capable, open source, and model-agnostic.
- Hermes' built-in learning loop is real: it writes and improves its own skills.
Where they stop
- Run ungoverned, OpenClaw has a rough record: 255+ security advisories reported, including a one-click token leak to remote code execution (CVE-2026-25253); 40,000+ instances found exposed online; and the ClawHavoc campaign, which planted 1,100+ malicious skills in its marketplace.
- Hermes is personal-scale and self-hosted, with no built-in multi-user governance. You own its security, governance and scaling.
- Skills aren't tested against your systems before they run.
Already using OpenClaw and Hermes?
Clone gives you the same open, any-model power with tested skills, scoped permissions and a full audit trail. Nothing to host, patch or secure yourself.
Bars show how fully each product covers the capability today. DIY: you'd build it yourself.
One shared computer, or a teammate with its own keys
Security has to live in the architecture. Clones get access denied by default, permissions set at the object level, and every action on the record.
- Every bot shares the computer's files, browser sessions and credentials.
- Deleting a bot leaves its files and browser sessions behind.
“Do not use separate Bots as a security boundary.”Grok Bot documentation, SpaceXAI
Questions buyers ask
Is Clone another AI model?
No. Clone runs on the leading models (GPT, Claude and Gemini) and uses the one that fits each task. What Clone adds is the teammate around the model: a role, your company's knowledge, skills it actually runs, the channels your people use, and governance on every action.
We already use ChatGPT or Claude. Why move to Clone?
Because your clones run on those same models and add everything a chat window lacks. Even for everyday writing, research and code, a clone starts from your company's knowledge, its memory of you and your sent email, where chat starts from a blank box. Clones also own outcomes: they work your inbox, phone line, CRM and meetings, and log every action for review. All of it on one Clone plan with usage included.
How is Clone different from Microsoft 365 Copilot?
Copilot makes people faster inside Word, Excel, Outlook and Teams. Clones do the work itself, inside Teams and Outlook and across every other system, including Salesforce, HubSpot and Google. They take calls, join meetings, and come on one flat plan with usage included, instead of a license plus metered agent charges.
How is Clone different from Grok Bot?
Grok's documentation says every bot on an account shares one cloud computer, including its files, browser sessions and credentials, and that separate bots shouldn't be used as a security boundary. Each clone has its own permissions, default-deny access and its own audit trail.
Couldn't we build this ourselves with OpenClaw, LangChain or n8n?
You could build an agent. Matching a clone means also building layered company knowledge, per-person memory, a skill pipeline that tests against live systems, object-level permissions, full audit, phone and meeting channels, and model routing, then securing and running all of it. Clone gives you the power of an open agent with those guardrails already built.
How does Clone pricing compare?
Clone plans are a flat monthly price with usage credits included. For example, Pro+ is $2,000 a month for up to 100 users, with $2,000 in credits included. Most alternatives charge a seat fee and then meter agent usage on top.
Which models does Clone use?
GPT, Claude and Gemini all run under the hood, so you're never locked into one provider.
How does Clone handle security?
Security is built into the architecture: access is denied by default on every route, permissions are set at the object level across orgs, departments, clones, skills and users, 100% of clone actions are logged, credentials are encrypted at rest, clones are pinned to specific skill versions, and each customer's data is kept separate. Clone runs on Google Cloud with Azure failover. Our SOC 2 Type II audit is in progress.
Bring the models you love.
Get the team that does the work.
Clones run on GPT, Claude and Gemini, in the tools you use, on one plan.